Application Of Ant Colony Optimization-based Support Vector Machine In Fault Diagnosis Of Transformer

Oil chromatography-line monitoring technique for early detection of latent faults of transformer has an important practical value. The traditional fault diagnosis methods by oil-gas component ratio are simple and practical, but the diagnostic accuracy is not high. Based on the analysis of traditional ratio methods and the common intelligent diagnostic methods, ant colony optimization-based support vector machine for fault diagnosis of transformer is put forward. The main contents are as follows:517sets of oil chromatographic data before accident are collected, and David trigonometry, four-ratio and the improved three-ratio method are used for diagnosis. The shortcomings of traditional diagnostic methods are demonstrated through the analysis of diagnostic results.Due to the large gap between the different types of gas content, data are standardized before being diagnosed. By comparing the diagnostic accuracy before and after data conversion, the influence of the data transformation on the diagnostic accuracy is analyzed. For the difference of gas type and detection precision of the current chromatograph, the correlation analysis method and distance separable criterion are used to select characteristic parameters and the influence of redundant and non-related parameters on the diagnosis accuracy is analyzed.The diagnostic accuracy of support vector machine is closely associated with the parameter selection. As the ant colony algorithm has a strong ability to find a better solution, ant colony optimization-based support vector machine is proposed. To demonstrate the feasibility and superiority of the method, we carry out a theoretical analysis and comparison with the widely used genetic algorithm, and describe the design of ant colony system in detail.A transformer fault diagnosis model based on ant colony optimization support vector machine is established based on the field data. Then, we focus on comparison of methods in this paper and genetic algorithm optimization support vector machine in fault diagnosis. Taking into account the impact of randomness of the optimization algorithm, we take multiple diagnostic tests and compare the results to increase the persuasiveness of the comparative results. Finally, the diagnostic results of improved three-ratio method and the method in this paper are compared, which prove the superiority of the ant colony optimization-based support vector machine in fault diagnosis.A transformer fault prediction model based on ant colony optimization support vector regression and time series is established based on the field data. Then, we focus on comparison of methods in this paper and grey prediction method in fault prediction. The simulation results show that transformer fault prediction based on ant colony optimization support vector regression model can be well applied to transformer oil dissolved gas content prediction, and the prediction accuracy is better than the gray prediction model, which has better generalization ability.